the invisible burden: goodwill and the cross-section of ... · returns after adjusted by all known...
Post on 17-Jan-2020
3 Views
Preview:
TRANSCRIPT
The Invisible Burden: Goodwill and the Cross-Section of Stock Returns*
Xin Liu† Chengxi Yin‡ Weinan Zheng§
First Draft: January 2018
This Version: January 2019
Abstract
We study the role of goodwill, an important form of intangible assets arising from past mergers and acquisitions, on asset pricing. We find that goodwill-to-sales strongly and negatively predicts the cross-section of stocks returns, especially among firms with cross-industry M&A histories and firms with overconfident CEOs. It remains as an economically and statistically significant predictor of stock returns after adjusted by all known factors. Our results suggest that goodwill-to-sales subsumes information on firm value, and stock markets underreact to this information because the fair value of goodwill is unobservable and hard to evaluate.
JEL Classification: G12, G14, G32, G34
Keywords: Goodwill, Return Predictability, Cash Flow, Underreaction, Market Inefficiency
* We thank Shiyang Huang, Dong Lou, Mike Adams, Xiaoran Ni, Vesa Pursiainen, Hanwen Sun, Chi-Yang Tsou, Hong
Xiang, Ru Xie, Tong Zhou, all seminar participants at University of Bath, The University of Hong Kong for helpful comments and suggestions. All errors are our own.
† Corresponding author. University of Bath, School of Management, Claverton Down, Bath, BA2 7AY, United Kingdom. Office 9.06, Wessex House. Phone: +44 (0) 1225 384297. E-mail: x.liu2@bath.ac.uk.
‡ China International Capital Corporation Limited, China World Office 2, 1 Jianguomenwai Avenue, Beijing, China. Phone: +86 (010) 6505 1166. E-mail: yinc1985@hotmail.com.
§ The University of Hong Kong, Faculty of Business and Economics, Pokfulam Road, Hong Kong. Office 1121, K.K.Leung Building. Phone: +852 3917 5343. E-mail: weinan.zheng@outlook.com.
The Invisible Burden: Goodwill and the Cross-Section of Stock Returns
First Draft: January 2018
This Version: January 2019
Abstract
We study the role of goodwill, an important form of intangible assets arising from past mergers and acquisitions, on asset pricing. We find that goodwill-to-sales strongly and negatively predicts the cross-section of stocks returns, especially among firms with cross-industry M&A histories and firms with overconfident CEOs. It remains as an economically and statistically significant predictor of stock returns after adjusted by all known factors. Our results suggest that goodwill-to-sales subsumes information on firm value, and stock markets underreact to this information because the fair value of goodwill is unobservable and hard to evaluate.
JEL Classification: G12, G14 G32, G34
Keywords: Goodwill, Return Predictability, Cash Flow, Underreaction, Market Inefficiency
1
1. Introduction
A firm’s stock price should reflect the value of both its tangible and intangible capitals.
While tangible capitals have been widely studied, intangible capitals have only received
growing attention in the recent decades due to the increasing importance in their economic
values. According to a report from Forbes, intangibles “have grown from filling 20% of
corporate balance sheets to 80%”, and they have become a crucial aspect determining the
market value of companies.1 The importance of taking intangible capitals into account to
evaluate firm values has also been emphasized in the recent literature, such as Chan,
Lakonishok, and Sougiannis (2001), Eisfeldt and Papanikolaou (2013), Belo, Lin, and Vitorino
(2014), and Peters and Taylor (2017). However, little attention has been paid to goodwill, which
is in fact the largest component of intangible capitals. As shown in Figure 1, the total dollar
value of goodwill in the U.S. stock markets has increased from $200 billion in 1989 to nearly
$5 trillion, consisting about 60% of total intangible assets.
[Figure 1 Here]
Unlike other intangibles, goodwill arises when a firm acquires another. It is measured as
the difference between the acquisition cost and the fair market value of the target’s identifiable
tangible and intangible net assets (Kieso, Weygandt, and Warfield, 2013). It may represent the
premium paid by the acquirer for the target’s resources (e.g., reputation, customer loyalty), as
well as the expected synergy generated by the business combination.2 Acquirers typically pay
1 https://www.forbes.com/sites/christopherskroupa/2017/11/01/how-intangible-assets-are-affecting-company-value-in-
the-stock-market/#5d772ef62b8e 2 For example, Hendriksen (1982, page 407) interprets goodwill as attitudes from employees, suppliers and customers.
Henning, Lewis, and Shaw (2000) adopt the synergy approach to identify the components of goodwill.
2
a huge premium, which is typically over 50% of the total acquisition price (KPMG, 2009).
However, a high premium does not necessarily guarantee optimistic future cash flows
(Malmendier and Tate, 2008). Often, acquirers paid substantial premium for targets but the
business combination end up underperforming than the initial expectation. If cash flows
generated from the combined firm are lower than expected, the balance sheet may give an
overly optimistic representation of a company’s financial health even when the income
statement does not justify this.3 This indicates that the fair value of goodwill may be mispriced.
Therefore, a high goodwill relative to cash flow could contain negative information on firm
value. In an efficient capital market, this negative information should be promptly incorporated
into stock prices. But as the “most intangible of intangible assets”,4 the fair value of goodwill
is not observable and hard to be accurately evaluated even for professional accountants, let
alone general investors. 5 Based on this argument, we hypothesize that stock markets
underreact to the information on firm value subsumed in goodwill to cash flow, and a high
goodwill relative to a firm’s cash flow can have a negative effect on future stock returns.
We find consistent evidence supporting this hypothesis. As our main test variable, we scale
goodwill by net sales from the beginning of the fiscal year (GTS) for all firms with a positive
goodwill account in their balance sheets.6 To take account of the variation of GTS in different
3 Having a high goodwill isn't a danger sign by itself. Older companies that have done many deals inevitably have lots of
goodwill. In addition, companies in high-tech and pharmaceutical sectors tend to have a high goodwill because they rely less on plants and machinery to make money.
4 Kieso, Donald E., Weygandt, Jerry J., Warfield, Terry D. Intermediate Accounting, 15th Edition. Hoboken, N.J. : Wiley, 2013. Print, page 659.
5 We elaborate the accounting of goodwill and its challenges in detail in Section 2.1. 6 We have tried different proxies for cash flow such as gross profits and net income. We have also considered other scaling
factors such as total assets and book value of assets. We have also considered using all scaling factors from the fiscal year-end. All these tests produce similar results and they are reported in Appendix Table A2.
3
industries, we compute industry-adjusted GTS (GTS_adj) as the difference between a firm’s
GTS and its industry mean GTS.7 A long-short portfolio that buys stocks from the lowest
GTS_adj decile and sells stocks from the highest GTS_adj decile earns a four-factor-adjusted
monthly return of 0.75% (t-statistic = 4.26). Robust results are also obtained after adjusted by
other common factor models, such as Fama-French (2015) five factors, Hou-Xue-Zhang (2015)
q-factors, Stambaugh-Yuan (2016)’s mispricing factors. 8 Fama-MacBeth regressions also
confirm our results.
To investigate whether this negative relation between goodwill-to-sales and the cross-
section of stock returns is driven by market underreaction, we examine the long-short portfolio
returns within five years after portfolio formation. The return predictability of GTS_adj decays
over time. The monthly four-factor alpha for the long-short portfolio monotonically decreases
from 0.75% (t-statistic = 4.26) in the first year to 0.31% (t-statistic = 2.38) in the third year. No
significant return patterns are obtained three years after the portfolio formation and the results
do not revert. These results support our hypothesis that market underreacts to goodwill-to-sales,
and stock price adjusts slowly to reflect the true value of the firm.
We further investigate the information content in goodwill-to-sales to pin down the
channel for underreaction. We conduct two sets of tests. First, we check whether high goodwill-
to-sales leads to high future goodwill impairment. Goodwill impairment is a reduction in
goodwill recorded on the income statement when goodwill's carrying value on the balance sheet
7 In the main analysis, we use Fama-French 38 industry classification to ensure cross-industry variations in GTS are well-
adjusted and there are sufficient stocks in each industry category. We have tried other industry adjustments and also GTS itself as the sorting variable and find consistent results. These results can be found in Appendix Table A3.
8 These results are reported in Appendix Table A4
4
exceeds its fair value. When an impairment occurs, the excess value of goodwill has to be
written off from the balance sheet, and the firm value drops. We scale goodwill impairment by
net sales from the beginning of the fiscal year (GITS). We conduct panel regressions with time
and industry fixed effects to investigate the relation between GTS and future GITS.9 Our result
shows that a one-standard-deviation increase in GTS leads to a 0.4% increase in next year’s
GITS (t-statistic = 2.89). For reference, the average GITS in our sample is 0.3%. Therefore,
this is a 133% increase in GITS relative to the sample mean. As our second test, we conduct
similar regressions to investigate the relation between goodwill-to-sales and future profitability.
We proxy profitability using return-on-assets (ROA), and our result show that a one-standard-
deviation increase in GTS leads to a 1.8% decrease in next year’s ROA (t-statistic = −4.69).
For reference, the average ROA in our sample is 3.6%. Therefore, this is a 50% drop in ROA
relative to the sample mean. These two tests both suggest that a high goodwill-to-sales contains
negative information about future firm value.
We further provide two empirical tests to support the information channel on the market
underreaction. The first empirical test is based on M&A histories. The degree of market
underreaction should depend on how complex the information is contained in goodwill-to-sales.
Existing studies have shown that market underreaction is more severe when the nature of the
information is more complex and more difficult to process (e.g. You and Zhang, 2009; Cohen
and Lou, 2012; Huang, 2015). Evaluating goodwill from a cross-industry M&A deal can be
substantially more complicated, because investors need to collect and analyze detailed
9 Because we have included industry fixed effects in the regressions, we do not need to adjust GTS within each industry.
Similar results are obtained if we run these regressions with GTS_adj instead of GTS.
5
information from two different industries, as well as to estimate synergies generated by the
combination of two different business segments. Therefore, our main results should be stronger
within the subsample of firms with cross-industry M&A histories. Consistent with this
prediction, we find that the negative relation between GTS_adj and subsequent stock returns
exists only among firms with cross-industry M&A histories. Within firms with cross-industry
M&A histories, the monthly four-factor alpha for a portfolio that longs stocks from the bottom
GTS_adj tercile and shorts stocks from the top GTS_adj tercile is 0.68% (t-statistic = 2.91).10
On the contrary, within firms with same-industry M&A histories, this alpha is 0.16% (t-statistic
= 0.57).
The second test to explore the information channel of market underreaction is based on
CEO overconfidence. Overconfident CEOs are more likely to make optimistic and less accurate
forecasts, delay loss recognition, adopt more aggressive accounting methods, conduct earnings
management, and engage in financial statement fraud, (Hillery and Hsu, 2011; Ahmed and
Duellman, 2013; Libby and Rennekamp, 2012; Schrand and Zechman, 2012; Bouwman, 2014;
Hribar and Yang, 2016; Banerjee, Humphery-Jenner, Nanda, and Tham, 2018). Therefore,
investors are difficult to judge the fair value of goodwill based on the biased information
released from financial reports. By this argument, we hypothesize that the negative relation
between goodwill-to-sales and subsequent stock returns should be stronger among firms with
overconfident CEOs. Indeed, we find that in the subsample with overconfident CEOs, the
monthly four-factor alpha for a portfolio that longs stocks from the bottom GTS_adj quintile
and shorts stocks from the top GTS_adj quintile is 0.69% per month (t-statistic = 3.93). In
10 We sort GTS_adj by tercile due to reduced sample size.
6
contrast, the same strategy only yields a four-factor alpha of 0.19% per month (t-statistic = 1.89)
in the subsample with non-overconfident CEOs.
Our paper has the following contributions to the literature. First, to our knowledge, we are
the first to document a negative relation between goodwill and subsequent stock returns. Our
long-short trading strategy produces an average monthly return that is not only statistically
significant, but also economically large, especially after adjusted by recent factor models
proposed by Fama and French (2015), Hou, Xue, and Zhang (2015), and Stambaugh and Yuan
(2016). Moreover, our results pass through the new statistic criteria proposed by Harvey, Liu,
and Zhu (2016) under all factor adjustments.11 Our strategy based on goodwill-to-sales is also
tradable. Unlike many anomalies, our results are not driven by small firms. We exclude stocks
with price below $5 and market capitalization below the bottom NYSE size decile. The median
NYSE size percentile in our sample is about 40%. Therefore, short selling stocks in the top
GTS_adj decile should be relatively easy.
Our paper also contributes to the literature studying the relation between intangible capitals
and the cross-section of stock returns. This literature can be further divided into three streams.
The first stream of literature studies intellectual properties (e.g., Chan, Lakonishok, and
Sougiannis, 2001; Gu, 2016; Hirshleifer, Hsu, and Li, 2013; Hirshleifer, Hsu, and Li, 2017).
They find that innovation intensity, efficiency, and originality positively predict the cross-
section of stock returns. The second stream of literature studies human capitals (e.g., Edmans,
2011; Eisfeldt and Papanikolaou, 2013). They find that firms with high employee satisfaction
11 Harvey, Liu, and Zhu (2016) suggest that new anomalies should be judged by a much higher statistic hurdle, with a t-
statistic greater than 3.0. Our average monthly long-short portfolio returns before and after risk adjustments all pass through this new hurdle, with t-statistics all greater than 3.5.
7
and more organization capital have higher returns. The third stream of literature analyzes the
effect of product branding on stock returns (e.g., Belo, Lin, and Vitorino, 2014; Lou, 2014; Hsu,
Li, Teoh, and Tseng, 2018). They document that firms with low brand capital investment rates,
more brand capital intensity, and more newly registered trademarks are associated with higher
future stock returns.
Our paper contributes to this literature as follows. First of all, even though intellectual
properties, human capitals, and product branding are all important factors for business
operations, goodwill is the largest component of intangible capitals and have significant
economic values, which is understudied in the asset pricing literature. Second, goodwill arises
from acquisition activities. It is fundamentally different in nature from other types of intangible
capitals. Therefore, the effect of goodwill on the cross-section of stock returns is also very
different compared to other intangibles. A high goodwill relative to cash flow contains negative
information on firm value. Since the fair value of goodwill is hard to evaluate, stock markets
underreact to this information. Thus, a high goodwill relative to cash flow negatively predicts
the cross-section of stock returns. This mechanism is different from other intangibles in the
literature.
Finally, our paper contributes to the M&A literature. Considerable evidence has shown
that the average announcement returns from making M&A is at best slightly positive, and
significantly negative in some studies (Bradley, Desai, and Kim, 1988; Roll, 1986; Morck,
Shleifer, and Vishny, 1990; Moeller, Schlingemann, and Stulz, 2005; Masulis, Wang, and Xie,
2007; Cai and Sevilir, 2012; among others). Our results suggest that a bad M&A deal has a
8
long-term impact on the combined firm after the announcement, especially for firms with cross-
industry M&A histories and overconfident CEOs.
The rest of the paper is organized as follows: Section 2 discusses the accounting of
goodwill, and describes our data and sample selection. Section 3 presents our main results.
Section 4 conducts further discussions. Section 5 concludes.
2. Data
2.1 The Accounting of Goodwill
Goodwill is among the most difficult-to-value assets on balance sheets. Under APB16,
SFAS141, and ASC805, goodwill should be booked using the purchase price method in which
it is calculated as the difference between the acquisition cost and the fair value of target firm’s
net assets. In order to calculate the fair value of the acquired unit, one needs to first identify its
liabilities, tangible assets and intangible assets that are separately identifiable from the business
combination at the first place. This identification step by itself could be complicated. After the
identification, one needs to figure out the fair market value for all these assets and liabilities,
which is more difficult because most assets do not have an active market. Therefore, one needs
to make assumptions on future cash flows and use valuation models to estimate the value of
these assets. These frictions together make it difficult to accurately value goodwill for any
business combination.
In order to improve goodwill accounting, U.S. FASB has changed rules for goodwill write-
off several times during the past two decades. However, it is still very challenging. From 1970
9
to 2001, APB17 required that goodwill should be amortized over a life time not to exceed 40
years. This requirement is controversial, because goodwill can be an asset with indefinite life
and its value might not decrease over time. Therefore, the information value of amortization is
very low as it is impossible to determine objectively the timeline over which amortization
should occur. Because of this issue, in 1995, U.S. FASB released SFAS121 which requires that
impairment should be taken when the goodwill value falls below the undiscounted future cash
flows. Since then, firms had been subject to both goodwill amortization and impairment until
2001. In 2001, SFAS142 superseded APB17 and SFAS121, and goodwill is no longer
amortized but only subject to an annual review for impairment. When testing whether a firm is
eligible for goodwill impairment, one needs to compare the carrying value of goodwill and the
implied fair value of goodwill. However, this impairment-only approach is also problematic:
(1) the annual impairment test is both costly and subjective; (2) the projections of future cash
flows from cash generating units is often too optimistic; (3) impairment losses tend to be
identified too late; (4) when an impairment loss is finally booked, the resulting information has
only weak confirmatory value for investors.
In order to reduce costs and efforts, FASB simplified the goodwill impairment test in 2017.
However, this simplification is also controversial. While simpler, the new procedure can be
less precise. As a result, it may “give rise to a goodwill impairment that is largely driven by
other assets in the reporting unit that are underwater but are not otherwise impaired under the
accounting literature”.12 All these regulatory changes indicate that evaluating goodwill is very
12 https://www.pwc.com/us/en/cfodirect/publications/in-the-loop/step-2-goodwill-impairment-test.html
10
challenging even for experienced professional accountants, let alone general investors.
The discretion in valuing goodwill has also left managers rooms to exploit. Watts (2003),
Hayn and Hughes (2006), Ramanna (2008), Ramanna and Watts (2012), Li and Sloan (2017),
and Glaum, Landsman and Wyrwa (2018) investigate the timing of goodwill impairments.
They document that firms with indicators of goodwill impairments tend to delay impairments
by taking advantage of discretionary use of valuation models.
2.2 Sample Construction
We start with all NYSE, AMEX and NASDAQ firms that are covered in Center for
Research in Security Prices (CRSP) and Compustat. We impose the following restrictions: 1)
a positive goodwill (Compustat data item: GDWL) at fiscal year-end; 2) a price-per-share larger
than $5; 3) a market capitalization higher than the bottom NYSE size decile. To mitigate
backfilling biases, a firm must be listed on Compustat for 2 years before it is included in the
dataset (Fama and French, 1992). Following Fama and French (1992), we match accounting
data for all fiscal year-ends in calendar year t−1 with the returns from July of year t to June of
year t+1 to ensure that the accounting variables are known before the returns they are used to
explain. We start all of our portfolio tests and regression analyses in the end of June 1989
because Compustat starts reporting goodwill in 1988. Our final sample covers 1989 to 2016.
Our main variable of interest, industry-adjusted goodwill-to-sales (GTS_adj) is
constructed as follows. We first compute goodwill-to-sales (GTS) as goodwill (GDWL) scaled
by total sales (SALE) from the beginning of the fiscal-year. To take account of the variation of
GTS in different industries, we compute industry-adjusted GTS (GTS_adj) as the difference
11
between GTS and the mean GTS from the same industry.13 We use Fama-French 38 industry
classifications to ensure cross-industry variations in GTS are well-adjusted and there are
sufficient stocks in each industry category. We require that each industry should have at least 3
firms to make this adjustment. In Appendix Tables A2 and A3, we consider alternative proxies
for goodwill and other industry adjustments, and obtain robust results.
Our control variables include: (1) Size, the market capitalization in billions of US dollars
at the end of each June; (2) Book-to-Market Ratio, defined as book equity over market equity.
We use book value from fiscal year-end t−1 and market value from December of year t−1; (3)
Momentum, defined as the cumulative returns from month t−12 to t−2; (4) Short Term Reversal,
defined as return of month t−1; (5) Idiosyncratic Volatility, defined as the monthly standard
deviation of the residuals from regressing daily returns on Fama-French (1993) three factors;
(6) Asset Growth, defined as the annual growth rate of total assets; (7) Gross Profit, defined as
the difference between total revenue and costs of goods sold, scaled by total assets; (8) Accruals,
calculated following Sloan (1996); (9) Net Stock Issuance, defined as the change in the natural
log of split-adjusted shares outstanding, following Pontiff and Woodgate (2008).
The first three columns in Panel A of Table 1 report the summary statistics for our full
sample. The mean of GTS is 0.261, and the median is 0.132.
[Table 1 Here]
Panel B of Table 1 reports the correlation coefficient matrix for our full sample. GTA has
13 The results are robust if we use industry median to adjust GTS.
12
no strong correlation with other well-documented firm characteristics.
We use the full sample for our main analyses, but we also divide this full sample based on
M&A histories and CEO overconfidence. We extract details on M&A deals from the Securities
Data Corporation (SDC) and divide our full sample based on M&A histories. We define a cross-
industry M&A as a deal in which the acquirer and the target belong to different Fama-French
38 industry classifications. Firms that have made at least one cross-industry M&A deal in the
past year are included in the subsample of cross-industry M&A. Firms that have only made
M&A deals within the same industry in the past year are included in the subsample of same-
industry M&A.
To capture CEO overconfidence, we follow Schrand and Zechman (2012) and bundle up 4
firm characteristics that are related to CEO overconfidence: (1) Excess Investment, defined as
capital expenditure scaled by total sales from the beginning of the fiscal-year; (2) Leverage
Ratio, defined as long-term plus short-term debt divided by total market value; (3) whether the
firm has outstanding preferred stocks or convertible debts; (4) whether the firm paid dividends
in the previous fiscal year. We rank the first two characteristics into 10 groups in ascending
orders separately. We assign rank 10 to a firm if it has outstanding preferred stocks or
convertible debts, and 1 otherwise. Similarly, we assign rank 10 to a firm if it did not pay
dividends in the previous fiscal year, and 1 otherwise. Then we compute the average rank from
the four characteristics as a proxy for CEO overconfidence. We require a stock to have all four
characteristics to compute this proxy. A firm is defined to have an overconfident CEO if this
average rank is above the top quintile. Non-overconfidence subsample contains firms with an
13
average rank below the bottom quintile.
Summary statistics for these subsamples are reported in Panel A of Table 1. Firms with
M&A histories in the past year (either cross-industry or same-industry) have higher GTS
compared to our full sample average. Firms with overconfident CEOs have higher GTS on
average, compared to firms with non-overconfident CEOs.
3. Results
3.1 Goodwill-to-sales and the Cross-section of Stock Returns
To test the relation between goodwill-to-sales and the cross-section of stock returns, we
first conduct univariate sorting based on industry-adjusted goodwill-to-sales. We match
GTS_adj for all fiscal year-ends in calendar year t−1 with monthly returns for July of year t to
June of year t+1. At the end of each June, stocks are sorted into decile portfolios based on
GTS_adj. We compute equal-weighted monthly excess returns for each of the decile portfolios,
as well as a long-short strategy which longs stocks in the bottom goodwill decile and shorts
stocks in the top goodwill decile. The average equal-weighted returns, together with Fama-
French (1993) three-factor alphas and Fama-French-Carhart (1997) four-factor alphas are
reported in Table 2. Our equal-weighted results are not driven by small stocks for the following
reasons. First, we have excluded stocks with price below $5 and stocks with market
capitalization below the NYSE bottom size decile. Second, we only focus our analysis on firms
with a positive goodwill, and firms with positive goodwill tend to be large firms. Indeed, the
median NYSE size percentile in our sample is about 40%. We report value-weighted results in
14
Appendix Table A1. They are very similar to the results we present here.
[Table 2 Here]
Table 2 presents a striking pattern that goodwill-to-sales negatively predicts subsequent
stock returns in the cross-section. For instance, the four-factor alpha for the bottom decile
portfolio is 0.42% (t-statistic = 2.86) per month. Four-factor alphas drop as GTS_adj increases.
The four-factor alpha becomes −0.33% (t-statistic = −2.41) for the top GTS_adj decile. A long-
short strategy which longs stocks in the bottom decile portfolio and shorts stocks in the top
decile portfolio earns a four-factor-adjusted return of 0.75% per month (t-statistic = 4.26).
Similar patterns are obtained using excess returns and other factor-adjusted returns. We plot
four-factor alphas for each decile portfolio in Figure 2 to demonstrate the negative relation
between industry-adjusted goodwill-to-sales and subsequent stock returns.
[Figure 2 Here]
Our strategy based on goodwill-to-sales is also tradable. Unlike many other anomalies,
our analyses focus on large firms. Therefore, short selling stocks in the top GTS_adj decile
should be relatively easy. In addition, Table 2 shows that our long-short portfolio returns come
from both the long leg and the short leg. Therefore, this trading strategy is feasible even for
investors with short-sale constraints.
We have considered alternative ways to construct our sorting variables. We have
considered: (1) using alternative proxies for cash flows such as gross profits and net income;
(2) replacing total sales by total assets and book assets; (3) scaling goodwill by accounting
15
variables at fiscal year-end instead of the beginning of the fiscal year; (4) using alternative
industry specifications to compute industry-adjusted goodwill-to-sales; (5) using goodwill-to-
sales without any industry adjustments. Section 4.3 outlines these robustness checks. Results
are reported in Appendix A2 and A3. All results are consistent with what we have reported here
in Table 2.
Our long-short trading strategy produces an average monthly return that is not only
statistically significant, but also economically large, especially after adjusted by recent factor
models proposed by Fama and French (2015), Hou, Xue, and Zhang (2015), and Stambaugh
and Yuan (2016). These results are reported in Appendix A4. Moreover, our results pass through
the new statistic criteria proposed by Harvey, Liu, and Zhu (2016) under all factor adjustments.
Harvey, Liu, and Zhu (2016) suggest that new anomalies should be judged by a much higher
statistic hurdle, with a t-statistic greater than 3.0. In Table 2, our average monthly long-short
portfolio returns before and after risk adjustments all pass through this new hurdle, with t-
statistics all greater than 4.0. In Appendix A4, all results have t-statistics greater than 3.5.
In Table 3, we report the factor loadings from the four-factor model for the bottom and
the top decile portfolios, as well as the long-short portfolio. Returns from the bottom and the
top decile portfolios are positively correlated with market, size, and value factors, but
negatively correlated with momentum factor. The last row of Table 3 shows the factor loadings
for the long-short portfolio. Returns from the long-short portfolio cannot be explained by the
four factors, as the loadings on these factors are low and mostly insignificant.
[Table 3 Here]
16
The primary advantage of the sorting strategy is that it offers a simple picture of how
average returns vary across the spectrum of goodwill-to-sales such that we do not impose a
functional form on the relations. However, we cannot control for other well-documented
characteristics that affect stock returns in the cross-section. Therefore, we conduct Fama and
Macbeth (1973) regressions to examine whether the results from the portfolio-level analysis
hold when other variables with return predictability are controlled for. Specifically, we conduct
the following regressions in each month:
Ri,t+1=αt+β1×GTS_adji,t+β2×Sizei,t+β3×BMi,t+β4×Momi,t+β5×Strevi,t+β6×IVOLi,t+β7×AGi,t
+β8×GPi,t+β9×NSi,t+β10×ACi,t+εi,t , (1)
where Ri,t+1 is the realized return on stock i in month t+1 (in percentage), GTS_adji,t is the
industry-adjusted goodwill for stock i in month t. We expect β1 to be significantly negative.
Control variables include: 1) Size, the natural log of total market capitalization at the end of
each June; 2) BM, book-to-market ratio, defined as book equity over market equity; 3) Mom,
momentum, defined as the cumulative returns from month t−12 to t−2; 4) Strev, short term
reversal, defined as return of month t−1; 5) IVOL, idiosyncratic volatility, defined as the
standard deviation of the residuals from regressions of daily returns on Fama-French (1993)
three factors at month t−1; 6) AG, asset growth, defined as the annual growth rate of total assets;
7) GP, growth profit, defined as the difference between total revenue and costs of goods sold,
scaled by total assets; 8) NS, net stock issuance, defined as the change in the natural log of
split-adjusted shares outstanding, following Pontiff and Woodgate (2008); 9) AC, accruals,
calculated following Sloan (1996).
17
Following Fama and French (1992), we match accounting data for all fiscal year-ends in
calendar year t−1 with the returns from July of year t to June of year t+1 to ensure that the
accounting variables are known before the returns they are used to explain. All variables are
winsorized at the 1% and 99% percentiles to eliminate the potential influence of outliers. We
conduct predictive regressions specified in Equation (1) every month, and report the time-series
average of the slope coefficients for our sample over the 324 months from July 1989 to June
2016. Newey-West adjusted t-statistics are reported in parentheses.
[Table 4 Here]
Table 4 shows that results from Fama-Macbeth regressions are consistent with our
univariate sorting results presented in Table 2. The coefficient on GTS_adj is significantly
negative. For example, in column (2), after controlling for well-known firm characteristics, the
coefficient on GTS_adj is −0.18 with a Newey-West adjusted t-statistic of −3.22. The
difference in mean GTS_adj between the top and bottom GTS_adj deciles is about 4.13. Thus,
the coefficient suggests that the difference in return between the bottom and top GTS_adj
deciles is approximately 0.74% (= −0.18×4.13) per month, which is similar in magnitude to
our sorting results. Overall, both univariate sorting and Fama-MacBeth regressions show a
negative and statistically significant relation between goodwill-to-sales and the cross-section
of stock returns.
We next examine the long-term return predictability of industry-adjusted goodwill-to-
sales. Specifically, we conduct the univariate sorting based on industry-adjusted goodwill as in
Table 2, and examine the long-short portfolio returns over the next five years. In Table 5, we
18
report the average monthly excess return and alphas for the long-short portfolio from July of
year t to June of year t+5.
[Table 5 Here]
Results reported in Table 5 show that the predictability power of goodwill-to-sales
gradually disappears over time. For instance, the four-factor alpha for the long-short portfolio
is 0.75% per month for the first year (t-statistic = 4.26), but drops to only 0.49% per month for
the second year (t-statistic = 3.26), and is only 0.31% per month for the third year (t-statistic =
2.38). No significant return predictability is found after the third year and the results do not
revert. This decreasing pattern over time suggests that stock markets underreact to goodwill-
to-sales and stock prices adjust slowly to reflect the true value of the firm.
3.2 The Information Content in Goodwill-to-sales
The previous subsection has documented that a high goodwill-to-sales is negatively
associated with subsequent stock returns in the cross-section. In this subsection, we present
evidence arguing that this negative return predictability is due to the fact that investors
underreact to the negative information on firm value associated with a high goodwill.
We first check whether goodwill-to-sales can positively predict goodwill impairment.
Goodwill impairment is a reduction in goodwill recorded on the income statement. It happens
when there is persuasive evidence that goodwill can no longer demonstrate financial results
that were expected from it. We define goodwill-impairment-to-sales (GITS) as the ratio of
goodwill impairment at fiscal year-end over total sales from the beginning of the fiscal year.
19
We conduct the following fixed effects panel regression:
GITSi,t+1=α+β1×GTSi,t+β2×GITSi,t+β3×Controli,t+Year FE+Industry FE+εi,t , (2)
where GITSi,t+1 is goodwill-impairment-to-sales for firm i in fiscal year t+1, GTSi,t is goodwill-
to-sales for firm i in fiscal year t. We expect β1 to be significantly positive. We control for
lagged goodwill-impairment-to-sales and other firm characteristics: 1) Size, the natural log of
total market capitalization at the end of each June; 2) BM, book-to-market ratio, defined as
book equity over market equity; 3) Mom, momentum, defined as the cumulative returns from
month t−12 to t−2; 4) AG, asset growth, defined as the annual growth rate of total assets; 5) NS,
net stock issuance, defined as the change in the natural log of split-adjusted shares outstanding,
following Pontiff and Woodgate (2008); 6) AC, accruals, calculated following Sloan (1996); 7)
NOA, net operating assets, the ratio of the difference between operating assets and operating
liabilities over total assets from the beginning of the fiscal year; and 8) IG, investment growth,
the annual growth rate of capital expenditure. Because we have controlled for industry fixed
effects, we do not adjust GTS and GITS by industry for this test. Results are very similar if we
use industry-adjusted GTS and GITS instead. All variables are winsorized at 1% and 99%. All
independent variables are standardized to have a mean of zero and a standard deviation of one.
Our analysis starts from 1996 because goodwill impairment is introduced by FASB in 1995.
Results are reported in Table 6.
[Table 6 Here]
In all specifications, the coefficient of GTS is positive and significant, indicating that a
higher goodwill-to-sales is associate with a higher goodwill impairment in the next fiscal year.
20
For instance, Column 4 shows that the coefficient on GTS is 0.004 (t-statistic = 2.89). This
indicates that a one standard-deviation increase in GTS will result in a 0.4% increase in the
goodwill impairment in the next year. For reference, the mean GITS in our sample is only 0.3%.
In other words, this is 133% increase relative to the mean.
As a second test, we check whether goodwill-to-sales negatively predicts profitability in
the next fiscal year. We take return-on-assets (ROA), defined as net income from fiscal year-
end over total asset from the beginning of the fiscal year, as a proxy for profitability. We regress
ROA in fiscal year t+1 on goodwill-to-sales from fiscal year t, and the same set of control
variables in Equation (2). All variables are winsorized at 1% and 99%. All independent
variables are standardized to have a mean of zero and a standard deviation of one.
ROAi,t+1=α+β1×GTSi,t+β2×GITSi,t+β3×Controli,t+Year FE+Industry FE+εi,t , (3)
[Table 7 Here]
Table 7 shows that a higher goodwill-to-sales is associated with a lower ROA in the
subsequent fiscal year. In all specifications, the coefficients for goodwill-to-sales is negative
and significant. For instance, in Column 4, the coefficient on GTS is −0.018 (t-statistic = −4.69).
This indicates that a one standard-deviation increase in GTS leads to a 1.8% decrease in ROA
in the next year. For reference, the mean ROA in our sample is 3.6%. In other words, this is
50% decrease relative to the mean.
Overall, in this subsection, we analyze why a high goodwill-to-sales is negatively
associated with subsequent stock returns in the cross-section. We present evidence showing
21
that investors underreact to the information associated with goodwill-to-sales, and stock prices
slowly adjust to reflect the true value of the firm. We present empirical results from M&A
histories and CEO overconfidence to further pin down this information channel in the next
section.
4. Further Discussion
The results documented so far suggest that stock markets underreact to the information on
firm value subsumed in goodwill-to-sales. We conduct further analyses to support this
information channel based on information complexity. Existing studies have shown that market
underreaction is more severe when the nature of the information is more complex and more
difficult to process (e.g. You and Zhang, 2009; Cohen and Lou, 2012; Huang, 2015). Following
this vein, we investigate whether our main results are stronger among firms with cross-industry
M&A histories (Section 4.1) and firms with overconfident CEOs (Section 4.2). In Section 4.3,
we discuss comprehensive robustness checks on our main results.
4.1 M&A Histories
Evaluating goodwill from a cross-industry M&A deal can be substantially more
complicated than evaluating goodwill from a same-industry M&A deal, because investors need
to collect and analyze detailed information from two industries, as well as to estimate synergies
generated by the combination of two different business segments. Therefore, we expect a
stronger negative relation between goodwill and subsequent stock returns for firms with cross-
industry M&A histories. To examine this prediction, we conduct independent double sorting
22
based on M&A histories and industry-adjusted goodwill-to-sales.14 We define a cross-industry
M&A as a deal in which the acquirer and the target belong to different Fama-French 38 industry
classifications. Firms that have made any cross-industry M&A deals in the past year are
included in the subsample of cross-industry M&A. Firms that have only made M&A deals
within the same industry in the past year are included in the subsample of same-industry M&A.
We sort industry-adjusted goodwill-to-sales by terciles independently.15 We report the equal-
weighted average monthly excess returns and alphas for the double sorting in Table 8.
[Table 8 Here]
Results in Table 8 show that the negative relation between goodwill-to-sales and
subsequent stock returns only exists within the subsample with cross-industry M&A histories.
Within firms with cross-industry M&A histories, a trading strategy which longs stocks in the
bottom industry-adjusted goodwill-to-sales tercile and shorts stocks in the top industry-
adjusted goodwill-to-sales tercile yields a four-factor alpha of 0.68% per month (t-statistic =
2.91). In contrast, within firms that only conduct same-industry M&A deals, this trading
strategy has a four-factor alpha of 0.16% (t-statistic = 0.57).16 Similar patterns are obtained
using excess returns and three-factor alphas.
These results further support our main finding that stock markets underreact to the
information subsumed in goodwill-to-sales. Because evaluating the information in goodwill-
14 Sequential double sorting produces similar results. 15 We sort goodwill-to-sales by terciles due to reduced sample size. 16 For reference, if we sort our full sample analyzed in Table 2 into terciles, a trading strategy which longs stocks in the
bottom industry-adjusted goodwill-to-sales tercile and shorts stocks in the top industry-adjusted goodwill-to-sales tercile has a four-factor alpha of 0.39% (t-statistic=1.63).
23
to-sales from past cross-industry M&A histories is a more complex and difficult job, the
underreaction effect is stronger for firms with cross-industry M&A histories.
4.2 CEO Overconfidence
To provide further evidence to the information channel for the underreaction effect, we
next investigate how CEO overconfidence affects the negative relation between goodwill-to-
sales and subsequent stock returns. Overconfident CEOs are more likely to make optimistic
and less accurate forecasts, delay loss recognition, adopt more aggressive accounting methods,
conduct earnings management, and engage in financial statement fraud (Hillery and Hsu, 2011;
Ahmed and Duellman, 2013; Libby and Rennekamp, 2012; Schrand and Zechman, 2012;
Bouwman, 2014; Hribar and Yang, 2016; Banerjee, Humphery-Jenner, Nanda, and Tham,
2018). Therefore, investors are difficult to judge the fair value of goodwill based on the biased
information released from financial reports. By this argument, we hypothesize that the negative
relation between GTS_adj and subsequent stock returns should be stronger among firms with
overconfident CEOs.
We conduct independent sort based on whether a firm has an overconfident CEO and
GTS_adj quintiles. The definition for overconfident CEOs is described in Section 2.2. We
report the equal-weighted average monthly excess returns and alphas for the double sorting in
Table 9.
[Table 9 Here]
Table 9 shows that the negative relation between industry-adjusted goodwill-to-sales and
24
subsequent stock returns is stronger within the subsample of overconfident CEOs. Among
firms with overconfident CEOs, a trading strategy which longs stocks in the bottom goodwill-
to-sales quintile and shorts stocks in the top goodwill-to-sales quintile yields a four-factor alpha
of 0.69% per month (t-statistic = 3.93). However, within firms with non-overconfident CEOs,
this trading strategy only yields a four-factor alpha of 0.19% (t-statistic = 1.89).17 Similar
patterns are obtained using excess returns and three-factor alphas.
These results further support our main finding that stock markets underreact to the
information subsumed in goodwill-to-sales. Because evaluating the information in goodwill-
to-sales from firms with overconfident CEOs is a more complex and difficult job, the
underreaction effect is stronger for firms with overconfident CEOs.
4.3. Robustness
In this section, we check the robustness of our main results. We first report value-weighted
portfolio returns in Appendix Table A1. The return patterns are similar with equal-weighted
results reported in Table 2. The long-short portfolio earns a four-factor adjusted return of 0.58%
per month (t-statistic = 3.65).
Next, we show that our results are robust across alternative definitions of the sorting
variables in Appendix Table A2. For our main analysis (also reported in column (1) of
Appendix Table A2), we use goodwill scaled by total sales from the beginning of the fiscal year
as our sorting variable. As a first test, in column (2), we scale goodwill by total sales at fiscal
17 For reference, if we sort our full sample analyzed in Table 2 into quintiles, a trading strategy which longs stocks in the
bottom industry-adjusted goodwill-to-sales quintile and shorts stocks in the top industry-adjusted goodwill-to-sales quintile has a four-factor alpha of 0.47% (t-statistic=4.12).
25
year-end. Results indicate that using total sales from fiscal year-end instead of the beginning
of the year does not alter our results.
Second, we consider total assets and book assets as denominators for the sorting variable
instead of total sales. In column (3), we scale goodwill by total assets from the beginning of
the fiscal year. In column (4), we scale goodwill by total assets at year-end. In column (5), we
scale goodwill by the book value of assets (defined as the sum of book value of equity and
liabilities) from the beginning of the fiscal year. In column (6), we scale goodwill by book
value of assets at fiscal year-end. All these results are consistent with what we have reported in
Table 2, suggesting that using total assets and book assets do not affect our main results.
We also consider alternative proxies for cash flows, i.e., gross profits and net income. In
column (7), we scale goodwill by gross profits from the beginning of the fiscal year. In column
(8), we scale goodwill by gross profits from fiscal year-end. In column (9), we scale goodwill
by net income from the beginning of the fiscal year. In column (10), we scale goodwill by net
income from the fiscal year-end. For all these alternative specifications, we find consistent
results that goodwill relative to cash flows is negatively associated with subsequent stock
returns in the cross-section.
We next examine whether our main results are sensitive to different industry adjustments.
For our main analysis (also reported in column (1) of Appendix Table A3), we use industry-
adjusted goodwill-to-sales based on Fama-French 38 industries. In order to show that our
results are not driven by industry adjustments, in column (2), we report sorting results by using
goodwill-to-sales without industry adjustments. We still find a significantly negative relation
26
between goodwill-to-sales and subsequent stock returns in the cross-section.
We consider other industry adjustments in Appendix Table A3. In column (3), we use
Fama-French 48 industries. In column (4), we use Fama-French 30 industries. In column (5),
we use Fama-French 17 industries. In column (6), we use Fama-French 5 industries. In column
(7), we use 1-digit SIC codes. In column (8), we use 2-digit SIC codes. For all these alternative
industry adjustments, we find consistent results that goodwill-to-sales is negatively associated
with subsequent stock returns in the cross-section.
Our main results presented in Table 2 are also robust after controlling for other factors. In
Appendix Table A4, we report adjusted returns based on (1) Fama-French (2015) five factors;
(2) Fama-French-Carhart six factors; (3) Hou-Xue-Zhang (2015) q-factors; (4) Stambaugh-
Yuan (2016)’s mispricing factors; and (5) DGTW benchmark portfolio returns. Our long-short
strategy is robust across all these alternative return adjustments.
Moreover, all of our results pass through the new statistic criteria proposed by Harvey,
Liu, and Zhu (2016). They suggest that new anomalies should be judged by a much higher
statistic hurdle, with a t-statistic greater than 3.0. All of our results from Table 2 and Appendix
Table A4 pass through this new hurdle, with t-statistics all greater than 3.5.
Finally, we conduct additional double sorts based on institutional ownership, idiosyncratic
volatility, and market capitalization. These results are reported in Appendix Table A5. We find
that the negative relation between goodwill-to-sales and subsequent stock return is much
stronger for firms with low institutional ownership, high idiosyncratic volatility, and low
market capitalization. These results are consistent with the information channel we have
27
discussed in Section 4.1 and Section 4.2. For stocks with low institutional ownership, high
idiosyncratic volatility, and low market capitalization, the fair value of goodwill is harder to
evaluate due to information asymmetry and uncertainty. Therefore, the negative relation
between goodwill-to-sales and subsequent stock returns are stronger for firms with low
institutional ownership, high idiosyncratic volatility, and low market capitalization.
5. Conclusion
In this paper, we study the asset pricing implications for the largest intangible asset, i.e.
goodwill. We argue that goodwill-to-cash-flow ratio contains information on firm value, and
investors underreact to this information because the fair value of goodwill is very hard to
evaluate. We conjecture that stocks with high goodwill-to-cash-flow ratio will experience lower
subsequent returns.
Consistent with our hypothesis, we show that stocks with high goodwill-to-sales
underperform stocks with low goodwill-to-sales, especially among firms with more complex
M&A histories and firms with overconfident CEOs. This negative relation is robust across
different industry adjustments, different proxies for cash flows, and different factor adjustments,
and it decays over time and disappears after three years since portfolio formation. Moreover, a
high goodwill-to-sale positively predicts future goodwill impairment and negatively predicts
future profitability. Overall, our results suggest that goodwill relative to cash flow does contain
information on firm value. Stock markets underreact to this information, therefore goodwill
relative to cash flow negatively predicts the cross-section of stock returns.
28
Our research has the following implications. First, investors should take into consideration
of goodwill-to-sales when evaluating stocks. They should avoid stocks with a high goodwill-
to-sale because this high level of goodwill cannot be well justified by firm performance, and
these stocks tend to experience lower returns. Second, when making acquisition decisions,
managers from acquirers should evaluate business combinations more rationally and accurately
to avoid a huge invisible burden, i.e., goodwill, on the balance sheet of the combined firm. Last,
regulators should also pay close attention to the goodwill-to-sales ratio, especially during M&A
booms. Stock markets with a huge aggregate goodwill could spell trouble for corporate
earnings and lead to painful write-offs.
29
References
Ahmed, A. S., & Duellman, S. (2013). Managerial overconfidence and accounting
conservatism. Journal of Accounting Research, 51(1), 1-30.
Banerjee, S., Humphrey-Jenner, M., Nanda, V., & Tham, M. (2018). Executive overconfidence
and securities class actions. Journal of Financial and Quantitative Analysis, forthcoming.
Belo, F., Lin, X., & Vitorino, M. A. (2014). Brand capital and firm value. Review of Economic
Dynamics, 17(1), 150-169.
Bouwman, C. H. (2014). Managerial optimism and earnings smoothing. Journal of Banking &
Finance, 41, 283-303.
Bradley, M., Desai, A., & Kim, E. H. (1988). Synergistic gains from corporate acquisitions and
their division between the stockholders of target and acquiring firms. Journal of Financial
Economics, 21(1), 3-40.
Cai, Y., & Sevilir, M. (2012). Board connections and M&A transactions. Journal of Financial
Economics, 103(2), 327-349.
Chan, L. K., Lakonishok, J., & Sougiannis, T. (2001). The stock market valuation of research
and development expenditures. The Journal of Finance, 56(6), 2431-2456.
Cohen, L., & Lou, D. (2012). Complicated firms. Journal of Financial Economics, 104(2),
383-400.
Edmans, A. 2011. Does the Stock Market Fully Value Intangibles? Employee Satisfaction and
Equity Prices. Journal of Financial Economics, 101: 621-640.
30
Eisfeldt, A. L., & Papanikolaou, D. (2013). Organization capital and the cross‐section of
expected returns. The Journal of Finance, 68(4), 1365-1406.
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. The Journal
of Finance, 47(2), 427-465.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and
bonds. Journal of Financial Economics, 33(1), 3-56.
Fama, & French. (2015). A five-factor asset pricing model. Journal of Financial
Economics, 116(1), 1-22.
Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. Journal
of Political Economy, 81(3), 607-636.
Glaum, M., Landsman, W. R., & Wyrwa, S. (2018). Goodwill Impairment: The Effects of
Public Enforcement and Monitoring by Institutional Investors. The Accounting Review,
forthcoming.
Gu, L. (2016). Product market competition, R&D investment, and stock returns. Journal of
Financial Economics, 119(2), 441-455.
Harvey, C. R., Liu, Y., & Zhu, H. (2016). … and the cross-section of expected returns. The
Review of Financial Studies, 29(1), 5-68.
Hayn, C., & Hughes, P. J. (2006). Leading indicators of goodwill impairment. Journal of
Accounting, Auditing & Finance, 21(3), 223-265.
Hendriksen, E. S. (1982), Accounting Theory, 4th edition., Homewood, Richard D. Irwin.
31
Henning, S. L., Lewis, B. L., Shaw, W. H. (2000), Valuation of the components of purchased
goodwill, Journal of Accounting Research, 38(2), 375-386.
Hirshleifer, D., Hsu, P. H., & Li, D. (2013). Innovative efficiency and stock returns. Journal of
Financial Economics, 107(3), 632-654.
Hirshleifer, D., Hsu, P. H., & Li, D. (2017). Innovative originality, profitability, and stock
returns. The Review of Financial Studies, 31(7), 2553-2605.
Hribar, P., and H. Yang. (2016). CEO Overconfidence and Management Forecasting.
Contemporary Accounting Research, 33(1), 204-227.
Hou, K., Xue, C., & Zhang, L. (2015). Digesting anomalies: An investment approach. The
Review of Financial Studies, 28(3), 650-705.
Hsu, P.H., Li, D., Teoh, S.H., and Tseng, K. (2019), Misvaluation of New Trademarks Available
at SSRN: https://ssrn.com/abstract=3224187.
Huang, X. (2015). Thinking Outside the Borders: Investors’ Underreaction to Foreign
Operations Information, The Review of Financial Studies, 28(1), 3109–3152.
Kieso, Donald E., Weygandt, Jerry J., Warfield, Terry D (2013). Intermediate Accounting.
Hoboken, N.J. : Wiley. Print.
KPMG. (2009). Intangible Assets and Goodwill in the context of Business Combinations.
https://www.consultancy.nl/media/KPMG%20-%20Intangible%20Assets%20and%20Goo
dwill-836.pdf
Li, K. K., & Sloan, R. G. (2017). Has goodwill accounting gone bad?. Review of Accounting
32
Studies, 22(2), 964-1003.
Libby, R., & Rennekamp, K. (2012). Self‐serving attribution bias, overconfidence, and the
issuance of management forecasts. Journal of Accounting Research, 50(1), 197-231.
Lou, D. (2014). Attracting investor attention through advertising. The Review of Financial
Studies, 27(6), 1797-1829.
Malmendier, U., & Tate, G. (2008). Who makes acquisitions? CEO overconfidence and the
market's reaction. Journal of Financial Economics, 89(1), 20-43.
Masulis, R. W., Wang, C., & Xie, F. (2007). Corporate governance and acquirer returns. The
Journal of Finance, 62(4), 1851-1889.
Morck, R., Shleifer, A., & Vishny, R. W. (1990). Do managerial objectives drive bad
acquisitions?. The Journal of Finance, 45(1), 31-48.
Moeller, S. B., Schlingemann, F. P., & Stulz, R. M. (2005). Wealth destruction on a massive
scale? A study of acquiring‐firm returns in the recent merger wave. The Journal of Finance,
60(2), 757-782.
Peters, R. H., & Taylor, L. A. (2017). Intangible capital and the investment-q relation. Journal
of Financial Economics, 123(2), 251-272.
Pontiff, J., & Woodgate, A. (2008). Share issuance and cross-sectional returns. The Journal of
Finance, 63(2), 921-945.
Ramanna, K. (2008). The implications of unverifiable fair-value accounting: Evidence from
the political economy of goodwill accounting. Journal of Accounting and Economics, 45(2),
33
253-281.
Ramanna, K., & Watts, R. L. (2012). Evidence on the use of unverifiable estimates in required
goodwill impairment. Review of Accounting Studies, 17(4), 749-780.
Roll, R. (1986). The hubris hypothesis of corporate takeovers. Journal of Business, 197-216.
Schrand, C. M., & Zechman, S. L. (2012). Executive overconfidence and the slippery slope to
financial misreporting. Journal of Accounting and Economics, 53(1), 311-329.
Sloan, R. G. (1996). Do stock prices fully reflect information in accruals and cash flows about
future earnings?. The Accounting Review, 71(3), 289-315.
Stambaugh, R. F., & Yuan, Y. (2016). Mispricing factors. The Review of Financial
Studies, 30(4), 1270-1315.
Watts, R. L. (2003). Conservatism in accounting part I: Explanations and implications.
Accounting Horizons, 17(3), 207-221.
You, H., & Zhang, X. J. (2009). Financial reporting complexity and investor underreaction to
10-K information. Review of Accounting Studies, 14(4), 559-586.
34
Table 1: Summary Statistics
This table reports descriptive statistics of firm characteristics for our samples. In Panel A, we report summaries from different samples. The first three columns report summary
statistics for our main sample. This sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-
share larger than $5, and a market capitalization higher than the bottom NYSE size decile. In Column 4-9, we select subsamples based on past mergers and acquisitions (M&A)
histories. We define a cross-industry M&A as a deal in which the acquirer and the target belong to different Fama-French 38 industry classifications. Firms that have made at
least one cross-industry M&A deal in the past year are included in the subsample of cross-industry M&A. Firms that have only made M&A deals within the same industry in
the past year are included in the subsample of same-industry M&A. In Column 10-15, we divide our full sample based on CEO overconfidence. Specifically, as suggested by
Schrand and Zechman (2012), we bundle up 4 firm characteristics that are related to CEO overconfidence: (1) Excess Investment, measured by capital expenditure scaled by
total sales from the beginning of the fiscal-year; (2) Leverage Ratio, defined as long-term plus short-term debt divided by total market value; (3) whether the firm has outstanding
preferred stocks or convertible debts; (4) whether the firm paid dividends in the previous fiscal year. We rank the first two characteristics into 10 groups on ascending order
separately. We assign rank 10 to a firm if it has outstanding preferred stocks or convertible debts and 1 otherwise. Similarly, we assign rank 10 to a firm if it did not pay dividends
in the previous year and 1 otherwise. Then we compute the average rank from the four characteristics as a proxy for CEO overconfidence. We require a stock to have all 4
characteristics to compute this proxy. A firm is defined to have an overconfident CEO if this average rank is above the top quintile. Non-overconfidence subsample contains
firms with an average rank below the bottom quintile. In Panel B, we report the correlation matrix for our full sample. Goodwill-to-Sales (GTS) is defined as goodwill divided
by total sales from the beginning of the fiscal-year. SIZE is the market capitalization in billions of US dollars at the end of each June. BM is book-to-market ratio, defined as
book equity over market equity. MOM is momentum, defined as the cumulative returns from month t−12 to t−2. STREV is short term reversal, defined as return of month t−1.
IVOL is idiosyncratic volatility, defined as the monthly standard deviation of the residuals from regressing daily returns on Fama-French (1993) three factors. AG is asset growth,
defined as the annual growth rate of total assets. GP is gross profit, defined as the difference between total revenue and costs of goods sold, scaled by total assets. AC is accruals
calculated following Sloan (1996). NS is net stock issuance, defined as the change in the natural log of split-adjusted shares outstanding, following Pontiff and Woodgate (2008).
Our samples cover 1989 to 2016 because Compustat starts reporting goodwill in 1988.
Panel A: Firm Characteristics
Full Sample Cross-Industry M&A Same-Industry M&A Overconfidence Non-overconfidence
Mean Median Std Mean Median Std Mean Median Std Mean Median Std Mean Median Std
GTS 0.261 0.132 0.348 0.358 0.236 0.383 0.402 0.250 0.442 0.335 0.175 0.420 0.205 0.106 0.281
SIZE 4.758 0.861 13.10 10.50 1.581 22.30 7.358 1.573 16.90 3.707 0.845 10.30 5.825 1.137 14.40
BM 0.575 0.489 0.385 0.492 0.418 0.334 0.485 0.408 0.342 0.652 0.551 0.442 0.533 0.461 0.339
MOM 0.138 0.091 0.438 0.092 0.061 0.434 0.110 0.068 0.447 0.139 0.077 0.494 0.134 0.106 0.344
STREV 0.009 0.008 0.115 0.006 0.007 0.121 0.008 0.006 0.123 0.008 0.007 0.127 0.010 0.009 0.095
IVOL 0.019 0.016 0.012 0.020 0.017 0.013 0.021 0.017 0.013 0.021 0.018 0.013 0.016 0.013 0.010
(Continued)
35
(Continued)
Panel A: Firm Characteristics
Full Sample Cross-Industry M&A Same-Industry M&A Overconfidence Non-overconfidence
Mean Median Std Mean Median Std Mean Median Std Mean Median Std Mean Median Std
AG 0.174 0.082 0.357 0.327 0.155 0.509 0.372 0.187 0.532 0.241 0.095 0.456 0.106 0.064 0.235
GP 0.328 0.301 0.225 0.349 0.322 0.178 0.363 0.332 0.198 0.267 0.24 0.175 0.345 0.331 0.248
AC 0.012 0.011 0.151 0.031 0.023 0.159 0.022 0.016 0.141 0.013 0.011 0.176 0.010 0.012 0.128
NS 0.040 0.006 0.141 0.082 0.011 0.195 0.097 0.017 0.203 0.067 0.012 0.173 0.016 0.001 0.111
Panel B: Correlation Matrix
GTS Size BM MOM STREV IVOL AG GP AC NS
GTS 1.000
SIZE 0.144 1.000
BM -0.021 -0.254 1.000
MOM -0.026 0.128 0.126 1.000
STREV -0.006 -0.020 -0.027 -0.141 1.000
IVOL -0.022 -0.399 0.029 -0.117 0.130 1.000
AG 0.223 -0.054 -0.152 -0.085 0.033 0.176 1.000
GP -0.278 -0.072 -0.291 0.006 0.011 0.046 -0.098 1.000
AC -0.021 -0.048 -0.024 -0.029 0.012 0.023 0.164 0.027 1.000
NS 0.190 -0.096 -0.038 -0.024 0.023 0.149 0.559 -0.135 -0.038 1.000
36
Table 2: Decile Portfolio Returns, 1989-2016
This table reports the equal-weighted average monthly excess returns and alphas to portfolios sorted on industry-adjusted goodwill-to-sales. The sample contains all common
shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5, and a market capitalization higher than the
bottom NYSE size decile. Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June
of year t+1 to ensure that the accounting variables are known before the returns they are used to explain. At the end of each June, we first compute goodwill-to-sales (GTS) as
the ratio of goodwill to total sales from the beginning of the fiscal year for all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal
year-end. Then, we compute industry-adjusted GTS (GTS_adj) as the difference between GTS and the mean GTS from the industry. We use Fama-French 38 industry
classifications, and require that each industry should have at least 3 firms to make this adjustment. After the industry adjustment, we exclude stocks with a share price less than
$5 or with a market capitalization below the bottom NYSE size decile. The rest of the stocks are sorted into decile portfolios based on GTS_adj. Portfolios are rebalanced at the
end of each June. The average difference in return between the bottom and the top decile portfolios are reported in the last column. We report excess returns, Fama-French three-
factor alphas, and Fama-French-Carhart four-factor alphas respectively. Newey-West adjusted t-statistics are reported in parentheses. The sample covers 1989 to 2016 because
Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in percentage.
Goodwill-to-Sales Deciles
Low 2 3 4 5 6 7 8 9 High Low − High
Excess returns 1.29*** 1.14*** 1.10*** 1.17*** 1.02*** 0.95*** 0.98*** 0.92*** 0.86*** 0.48 0.81*** (4.37) (4.39) (3.94) (4.19) (3.43) (3.18) (3.28) (3.26) (3.03) (1.52) (4.43)
Three-factor alpha 0.30* 0.15 0.07 0.16 -0.04 -0.11 -0.09 -0.12 -0.20* -0.56*** 0.86*** (1.92) (1.27) (0.47) (1.47) (-0.38) (-0.93) (-0.73) (-0.85) (-1.80) (-3.34) (4.33)
Four-Factor alpha 0.42*** 0.24** 0.22 0.27*** 0.10 0.07 0.10 0.04 -0.06 -0.33** 0.75***
(2.86) (2.28) (1.50) (2.70) (0.99) (0.77) (0.94) (0.31) (-0.61) (-2.41) (4.26)
37
Table 3: Factor Loading
This table presents the results of time-series regressions of the bottom and the top decile portfolio returns sorted by
industry-adjusted goodwill-to-sales, as well as the long-short portfolio returns on Fama-French (1993) three factors and
Carhart (1997)’s momentum factor. All portfolio returns are equal-weighted. The sample contains all common shares
traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5,
and a market capitalization higher than the bottom NYSE size decile. Following Fama and French (1992), we match
accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June of year t+1 to
ensure that the accounting variables are known before the returns they are used to explain. At the end of each June, we
first compute goodwill-to-sales (GTS) as the ratio of goodwill to total sales from the beginning of the fiscal year for all
common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. Then, we
compute industry-adjusted GTS (GTS_adj) as the difference between GTS and the mean GTS from the industry. We use
Fama-French 38 industry classifications, and require that each industry should have at least 3 firms to make this
adjustment. After the industry-adjustment, we exclude stocks with a share price less than $5 or with a market
capitalization below the bottom NYSE size decile. The rest of the stocks are sorted into decile portfolios based on
GTS_adj. Portfolios are rebalanced at the end of each June. Newey-West adjusted t-statistics are reported in parentheses.
The sample covers 1989 to 2016 because Compustat starts reporting goodwill in 1988. *, **, and *** indicate
significance at 10%, 5%, and 1%, respectively
Alpha MKTRF SMB HML UMD
Low 0.42*** 1.02*** 0.54*** 0.11 -0.14* (2.86) (24.00) (6.30) (1.48) (-1.89)
High -0.33** 1.03*** 0.61*** 0.14** -0.26*** (-2.41) (28.55) (6.33) (2.23) (-6.58)
Low − High 0.75*** -0.01 -0.07 -0.03 0.12*
(4.26) (-0.29) (-1.32) (-0.46) (1.80)
38
Table 4: Fama-MacBeth Regression, 1989-2016
This table reports the average coefficients and their respective Newey-West adjusted t-statistics from monthly firm-level
cross-sectional regressions of the return in that month on lagged variables including industry-adjusted goodwill-to-sales
and other control variables. The sample contains all common shares traded in NYSE, AMEX and NASDAQ that have
a positive goodwill at fiscal year-end, a price-per-share larger than $5, and a market capitalization higher than the bottom
NYSE size decile. Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar
year t−1 with the returns for July of year t to June of year t+1 to ensure that the accounting variables are known before
the returns they are used to explain. We first compute goodwill-to-sales (GTS) as goodwill divided by total sales from
the beginning of the fiscal-year for all common shares traded in NYSE, AMEX and NASDAQ that have a positive
goodwill at fiscal year-end. GTS_adj is the industry adjusted GTS, defined as the difference between GTS and the mean
GTS from the industry. We use Fama-French 38 industry classifications, and require that each industry should have at
least 3 firms to make this adjustment. SIZE is the natural log of total market capitalization at the end of each June. BM
is book-to-market ratio, defined as book equity over market equity. MOM is momentum, defined as the cumulative
returns from month t−12 to t−2. STREV is short term reversal, defined as return of month t−1. IVOL is idiosyncratic
volatility, defined as the standard deviation of the residuals from regressing daily returns on Fama-French (1993) three
factors at month t−1. GP is gross profit, defined as the difference between total revenue and costs of goods sold, scaled
by total assets. AG is asset growth, defined as the annual growth rate of total assets. AC is accruals calculated following
Sloan (1996). NS is net stock issuance, defined as the change in the natural log of split-adjusted shares outstanding,
following Pontiff and Woodgate (2008). Our samples cover 1989 to 2016 because Compustat starts reporting goodwill
in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively.
Variable (1) (2)
GTS_adj -0.27*** -0.18*** (-3.81) (-3.22)
SIZE -0.06* (-1.66)
BM 0.06 (0.36)
MOM 0.34 (1.01)
STREV -2.86*** (-5.17)
IVOL -17.75*** (-2.71)
AG -0.18* (-1.66)
GP 0.44**
(2.15)
NS -0.92*** (-3.36)
AC -0.14 (-0.66)
N 324 324
R-squared 0.003 0.061
39
Table 5: Long-term Return Predictability
This table reports the equal-weighted average monthly returns and alphas to a trading strategy
which buys stocks in the bottom goodwill-to-sales decile and shorts stocks in the top goodwill-
to-sales decile. We sort stocks into decile portfolios at the end of each June and track monthly
returns in the next five years. The sample contains all common shares traded in NYSE, AMEX
and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5,
and a market capitalization higher than the bottom NYSE size decile. Following Fama and
French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the
returns for July of year t to June of year t+1 to ensure that the accounting variables are known
before the returns they are used to explain. At the end of each June, we first compute goodwill-
to-sales (GTS) as the ratio of goodwill to total sales from the beginning of the fiscal year for all
common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal
year-end. Then, we compute industry-adjusted GTS (GTS_adj) as the difference between GTS
and the mean GTS from the industry. We use Fama-French 38 industry classifications, and
require that each industry should have at least 3 firms to make this adjustment. After the
industry-adjustment, we exclude stocks with a share price less than $5 or with a market
capitalization below the bottom NYSE size decile. The rest of the stocks are sorted into decile
portfolios based on GTS_adj. Portfolios are rebalanced at the end of each June. We compute
average monthly returns in the next five years after the portfolio formation. We report average
returns, Fama-French 3-factor alphas, and Fama-French-Carhart 4-factor alphas for the long-
short portfolio, respectively. Newey-West adjusted t-statistics are reported in parentheses. The
sample covers 1989 to 2016 because Compustat starts reporting goodwill in 1988. *, **, and
*** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in
percentage.
Year
t+1
Year
t+2
Year
t+3
Year
t+4
Year
t+5
Low − High 0.81*** 0.49*** 0.37*** 0.12 0.13 (4.43) (3.46) (2.93) (0.87) (0.86)
Three-factor alpha 0.86*** 0.59*** 0.40*** 0.16 0.12 (4.33) (3.94) (3.09) (1.08) (0.80)
Four-Factor alpha 0.75*** 0.49*** 0.31** 0.15 0.12 (4.26) (3.26) (2.38) (1.04) (0.76)
40
Table 6: Regression of Goodwill Impairment on Goodwill-to-Sales
This table reports panel regressions of goodwill impairment on lagged variables including goodwill-to-sales and other
control variables. The dependent variable, goodwill-impairment-to-sales (GITS), is the ratio of goodwill impairment in
fiscal year t+1 over total sales from the beginning of fiscal year t+1. Independent variables include: goodwill-to-sales
(GTS), the ratio of goodwill over total sales from the beginning of the fiscal-year; lagged goodwill-impairment-to-sales
from fiscal year t; size, the natural log of total market capitalization; book-to-market ratio (BM), book equity over market
equity; momentum (MOM), cumulative returns from month t−12 to t−2; asset growth (AG), the annual growth rate of
total assets; net stock issuance (NS), the change in the natural log of split-adjusted shares outstanding, following Pontiff
and Woodgate (2008); accruals (AC) calculated following Sloan (1996); net operating assets (NOA), the ratio of the
difference between operating assets and operating liabilities over total assets from the beginning of the fiscal-year;
investment growth (IG), the annual growth rate of capital expenditure. All variables are winsorized at 1% and 99%. All
independent variables are standardized to have a mean of zero and a standard deviation of one. In the first two columns,
we control for time-fixed effects and cluster standard errors by industry; in the third column, we control for time and
industry fixed-effects and cluster standard errors by industry; in the last column, we control for time and industry fixed-
effects and double cluster standard errors by time and industry. We use Fama-French 38 industry classifications. The
sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-
end, a price-per-share larger than $5, a market capitalization higher than the bottom NYSE size decile, a positive
goodwill impairment, as well as sufficient data to compute control variables. The sample covers 1996 to 2016 because
Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively.
Variable (1) (2) (3) (4)
GTS 0.003*** 0.004*** 0.004*** 0.004***
(2.97) (3.72) (3.40) (2.89)
GITS(t) 0.001*** 0.001*** 0.001**
(3.34) (3.18) (2.26)
SIZE -0.001*** -0.001*** -0.001***
(-4.88) (-4.13) (-2.69)
BM 0.002*** 0.002*** 0.002***
(3.74) (3.92) (3.28)
MOM
-0.001*** -0.001*** -0.001**
(-6.39) (-5.81) (-2.28)
AG -0.000 -0.000 -0.000
(-0.48) (-0.63) (-0.61)
NS 0.001** 0.001** 0.001
(2.31) (1.99) (1.55)
AC 0.001 0.001 0.001
(1.09) (1.10) (1.19)
NOA -0.001* -0.000 -0.000 (-1.71) (-1.14) (-1.09)
IG -0.000 -0.000 -0.000
(-0.09) (-0.04) (-0.04)
Year FE Yes Yes Yes Yes
Industry FE No No Yes Yes
Cluster Industry Industry Industry Industry, Year
Observations 37,528 28,716 28,716 28,716
R-squared 0.017 0.027 0.029 0.029
41
Table 7: Regression of ROA on Goodwill-to-Sales
This table reports panel regressions of return-on-assets on lagged variables including goodwill-to-sales and other control
variables. The dependent variable, return-on-assets (ROA), is net income over total assets from fiscal year t+1.
Independent variables include: goodwill-to-sales (GTS), the ratio of goodwill over total sales from fiscal-year t;
goodwill-impairment-to-sales from fiscal year t; size, the natural log of total market capitalization; book-to-market ratio
(BM), book equity over market equity; momentum (MOM), cumulative returns from month t−12 to t−2; asset growth
(AG), the growth rate of total assets; net stock issuance (NS), the change in the natural log of split-adjusted shares
outstanding, following Pontiff and Woodgate (2008); accruals (AC) calculated following Sloan (1996); net operating
assets (NOA), the ratio of the difference between operating assets and operating liabilities over total assets from the
beginning of the fiscal-year; investment Growth (IG), the annual growth rate of capital expenditure. All variables are
winsorized at 1% and 99%. All independent variables are standardized to have a mean of zero and a standard deviation
of one. In the first two columns, we control for time-fixed effects and cluster standard errors by industry; in the third
column, we control for time and industry fixed-effects and cluster standard errors by industry; in the last column, we
control for time and industry fixed-effects and double cluster standard errors by time and industry. We use Fama-French
38 industry classifications. The sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a
positive goodwill at fiscal year-end, a price-per-share larger than $5, a market capitalization higher than the bottom
NYSE size decile, a positive goodwill impairment, as well as sufficient data to compute control variables. The sample
covers 1989 to 2016 because Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%,
5%, and 1%, respectively.
Variable (1) (2) (3) (4)
GTS -0.015*** -0.017*** -0.018*** -0.018***
(-5.20) (-5.39) (-4.61) (-4.69)
GITS -0.005*** -0.005*** -0.005***
(-8.06) (-8.32) (-6.15)
SIZE 0.011*** 0.011*** 0.011***
(5.27) (4.09) (3.98)
BM -0.027*** -0.027*** -0.027***
(-25.36) (-25.82) (-12.71)
MOM
0.009*** 0.009*** 0.009***
(8.05) (7.92) (4.52)
AG 0.001 0.001 0.001
(0.92) (0.94) (0.92)
NS -0.002 -0.002 -0.002
(-1.39) (-1.31) (-1.15)
AC -0.015*** -0.015*** -0.015***
(-7.25) (-6.79) (-6.78)
NOA 0.015*** 0.015*** 0.015***
(3.14) (2.84) (2.83)
IG -0.001* -0.001* -0.001
(-1.82) (-1.74) (-1.43)
Year FE Yes Yes Yes Yes
Industry FE No No Yes Yes
Cluster Industry Industry Industry Industry, Year
Observations 35,481 28,720 28,720 28,720
R-squared 0.044 0.177 0.187 0.187
42
Table 8: Independent Double Sorting Based on Cross-Industry M&A, 1989-2016
This table reports the equal-weighted average monthly excess returns and alphas to portfolios double sorted on industry-adjusted goodwill-to-sales and mergers
and acquisition (M&A) histories. We define a cross-industry M&A as a deal in which the acquirer and the target belong to different Fama-French 38 industry
classifications. Firms that have made any cross-industry M&A deals in the past year are included in the subsample of cross-industry M&A. Firms that have
only made M&A deals within the same industry in the past year are included in the subsample of same-industry M&A. We sort industry-adjusted goodwill
independently based on whole sample terciles. The sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at
fiscal year-end, a price-per-share larger than $5, a market capitalization higher than the bottom NYSE size decile, and at least one M&A deal from the last year.
Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June of year t+1
to ensure that the accounting variables are known before the returns they are used to explain. At the end of each June, we first compute goodwill-to-sales (GTS)
as the ratio of goodwill to total sales from the beginning of the fiscal year for all common shares traded in NYSE, AMEX and NASDAQ that have a positive
goodwill at fiscal year-end. Then, we compute industry-adjusted GTS (GTS_adj) as the difference between GTS and the mean GTS from the industry. We use
Fama-French 38 industry classifications, and require that each industry should have at least 3 firms to make this adjustment. After the industry-adjustment, we
exclude stocks with a share price less than $5 or with a market capitalization below the bottom NYSE size decile. The rest of the stocks with at least one
acquisition in the previous year are double sorted into terciles portfolios based on GTS_adj and the M&A histories independently. Portfolios are rebalanced at
the end of each June. We report excess returns, Fama-French three-factor alphas, and Fama-French-Carhart four-factor alphas, respectively. Newey-West
adjusted t-statistics are reported in parentheses. The sample covers 1989 to 2016 because Compustat starts reporting goodwill in 1988. *, **, and *** indicate
significance at 10%, 5%, and 1%, respectively. All returns and alphas are in percentage.
Cross-industry M&A Same-industry M&A
Low Medium High Low − High Low Medium High Low − High
Excess return 1.14*** 0.87*** 0.51*** 0.63*** 0.98*** 0.99*** 0.83** 0.15 (3.56) (2.91) (2.88) (2.88) (3.31) (3.34) (2.17) (0.56)
Three-factor alpha 0.19 -0.16 -0.56*** 0.75*** 0.00 -0.03 -0.20 0.20 (0.87) (-0.67) (-2.89) (3.38) (0.05) (-0.19) (-0.77) (0.77)
Four-factor alpha 0.39* 0.04 -0.29* 0.68*** 0.17 0.12 0.01 0.16 (1.80) (0.19) (-1.81) (2.91) (0.89) (0.65) (0.02) (0.57)
43
Table 9: Independent Double Sorting Based on CEO Overconfidence, 1989-2016
This table reports the equal-weighted average monthly excess returns and alphas to portfolios double sorted on industry-adjusted goodwill-to-sales and CEO overconfidence.
As suggested by Schrand and Zechman (2012), we bundle up 4 firm characteristics that are related to CEO overconfidence: (1) Excess Investment, measured by capital
expenditure scaled by total sales from the beginning of the fiscal-year; (2) Leverage Ratio, defined as long-term plus short-term debt divided by total market value; (3) whether
the firm has outstanding preferred stocks or convertible debts; (4) whether the firm paid dividends in the previous fiscal year. We rank the first two characteristics into 10 groups
in the ascending order separately. We assign rank 10 to a firm if it has outstanding preferred stocks or convertible debts and 1 otherwise. Similarly, we assign rank 10 to a firm
if it did not pay dividends in the previous year and 1 otherwise. Then we compute the average rank from the four characteristics as a proxy for CEO overconfidence. We require
a stock to have all 4 characteristics to compute this proxy. A firm is defined to have an overconfident CEO if this average rank is above the top quintile. Non-overconfidence
subsample contains firms with an average rank below the bottom quintile. The sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a positive
goodwill at fiscal year-end, a price-per-share larger than $5, a market capitalization higher than the bottom NYSE size decile, and a non-missing proxy for CEO overconfidence.
Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June of year t+1 to ensure that
the accounting variables are known before the returns they are used to explain. At the end of each June, we first compute goodwill-to-sales (GTS) as the ratio of goodwill to
total sales from the beginning of the fiscal year for all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. Then, we compute
industry-adjusted GTS (GTS_adj) as the difference between GTS and the mean GTS from the industry. We use Fama-French 38 industry classifications, and require that each
industry should have at least 3 firms to make this adjustment. After the industry adjustment, we exclude stocks with a share price less than $5 or with a market capitalization
below the bottom NYSE size decile. The rest of the stocks with non-missing proxy for CEF overconfidence are double sorted into quintile portfolios based on GTS_adj and the
average rank of CEO overconfidence independently. Portfolios are rebalanced at the end of each June. We report excess returns, Fama-French three-factor alphas, and Fama-
French-Carhart four-factor alphas, respectively. Newey-West adjusted t-statistics are reported in parentheses. The sample covers 1989 to 2016 because Compustat starts reporting
goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in percentage.
Overconfidence Sample Non-overconfidence Sample
Low 2 3 4 High Low-High Low 2 3 4 High Low-High
Excess return 1.17*** 0.99*** 0.88** 1.03*** 0.40 0.77*** 1.15*** 1.17*** 1.05*** 1.02*** 0.97*** 0.18* (3.51) (2.95) (2.28) (3.04) (1.15) (4.22) (5.03) (4.68) (3.87) (3.75) (3.82) (1.85)
Three-factor alpha 0.08 -0.13 -0.29* -0.15 -0.70*** 0.78*** 0.27** 0.23* 0.07 0.04 0.02 0.25** (0.51) (-0.81) (-1.68) (-1.05) (-3.88) (4.09) (2.08) (1.94) (0.57) (0.28) (0.19) (2.45)
Four-factor alpha 0.25 0.16 -0.11 0.05 -0.44*** 0.69*** 0.31** 0.29*** 0.18* 0.16 0.12 0.19* (1.61) (1.26) (-0.74) (0.31) (-3.17) (3.93) (2.47) (2.71) (1.69) (1.32) (1.05) (1.89)
44
Figure 1. Time-series of Aggregate Goodwill and Aggregate Intangible Assets
This figure shows the time-series of the aggregate goodwill and aggregate intangible assets value across all U.S.
listed firms on Compustat. The sample period is from 1989 to 2016. The unit of measure is trillion dollars.
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
Trill
ions
of D
olla
rs
Year
Goodwill (Trillions of Dollars) Intangible Assets (Trillion of Dollars)
45
Figure 2: Average Monthly Four-factor Alpha for Decile Portfolios
This figure shows Fama-French-Carhart four-factor alphas for each decile portfolio sorted on
industry-adjusted goodwill-to-sales. The sample contains all common shares traded in NYSE,
AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than
$5, and a market capitalization higher than the bottom NYSE size decile. Following Fama and
French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the returns
for July of year t to June of year t+1 to ensure that the accounting variables are known before the
returns they are used to explain. At the end of each June, we first compute goodwill-to-sales (GTS)
as the ratio of goodwill to total sales from the beginning of the fiscal year for all common shares
traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. Then, we
compute industry-adjusted GTS (GTS_adj) as the difference between GTS and the mean GTS from
the industry. We use Fama-French 38 industry classifications, and require that each industry should
have at least 3 firms to make this adjustment. After the industry adjustment, we exclude stocks with
a share price less than $5 or with a market capitalization below the bottom NYSE size decile. The
rest of the stocks are sorted into decile portfolios based on GTS_adj. Portfolios are rebalanced at the
end of each June.
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Low 2 3 4 5 6 7 8 9 High
Ave
rage
Mon
thly
Fou
r-fa
ctor
Alp
ha (%
)
Portfolios Sorted by Goodwill-to-Sales
46
Table A1: Value-weighted Decile Portfolio Returns, 1989-2016
This table reports the value-weighted average monthly excess returns and alphas to portfolios sorted on industry-adjusted goodwill-to-sales. The sample contains
all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5, and a market
capitalization higher than the bottom NYSE size decile. Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar year
t−1 with the returns for July of year t to June of year t+1 to ensure that the accounting variables are known before the returns they are used to explain. At the
end of each June, we first compute goodwill-to-sales (GTS) as the ratio of goodwill to total sales from the beginning of the fiscal year for all common shares
traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. Then, we compute industry-adjusted GTS (GTS_adj) as the difference
between GTS and the mean GTS from the industry. We use Fama-French 38 industry classifications, and require that each industry should have at least 3 firms
to make this adjustment. After the industry-adjustment, we exclude stocks with a share price less than $5 or with a market capitalization below the bottom
NYSE size decile. The rest of the stocks are sorted into decile portfolios based on GTS_adj. Portfolios are rebalanced at the end of each June. The average
difference in return between the bottom and the top decile portfolios are reported in the last column. We report excess returns, Fama-French three-factor alphas,
and Fama-French-Carhart four-factor alphas, respectively. Newey-West adjusted t-statistics are reported in parentheses. The sample covers 1989 to 2016 because
Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in percentage.
Goodwill-to-Sales Deciles
Low 2 3 4 5 6 7 8 9 High Low − High
Excess returns 1.06*** 0.95*** 0.87*** 1.15*** 0.95*** 0.85*** 0.85*** 0.97*** 0.60** 0.59** 0.47*** (3.70) (3.76) (2.72) (3.59) (3.36) (3.07) (3.38) (3.60) (2.05) (2.06) (2.67)
Three-factor alpha 0.27* 0.16 -0.07 0.25 0.00 -0.07 0.01 0.03 -0.37*** -0.33*** 0.60***
(1.69) (1.08) (-0.53) (1.59) (0.03) (-0.6) (0.11) (0.28) (-3.12) (-2.78) (3.57)
Four-factor alpha 0.31** 0.10 0.02 0.26* 0.10 0.02 0.04 0.10 -0.32*** -0.27** 0.58*** (2.07) (0.69) (0.18) (1.72) (0.86) (0.28) (0.34) (0.92) (-2.80) (-2.55) (3.65)
47
Table A2: Alternative Goodwill-to-Cash-Flow Proxies
This table reports the equal-weighted average monthly excess returns and alphas to portfolios sorted on different proxies of industry-adjusted goodwill-to-cash-flow. In
column (1), we scale goodwill by total sales from the beginning of the fiscal year. In column (2), we scale goodwill by total sales at fiscal year-end. In column (3), we
scale goodwill by total assets from the beginning of the fiscal year. In column (4), we scale goodwill by total assets at year-end. In column (5), we scale goodwill by the
book value of assets from the beginning of the fiscal year. In column (6), we scale goodwill by book value of assets at fiscal year-end. In column (7), we scale goodwill
by the gross profits from the beginning of the fiscal year. In column (8), we scale goodwill by the gross profits from the end of the fiscal year. In column (9), we scale
goodwill by net income from the beginning of the fiscal year. In column (10), we scale goodwill by net income from the end of the fiscal year. The sample contains all
common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5, and a market capitalization
higher than the bottom NYSE size decile. Following Fama and French (1992), we match accounting data for all fiscal year-ends in calendar year t−1 with the returns
for July of year t to June of year t+1 to ensure that the accounting variables are known before the returns they are used to explain. At the end of each June, we first
compute the goodwill proxies for all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. We conduct industry
adjustments as the difference between goodwill proxies and their means from the industry. We use Fama-French 38 industry classifications, and require that each
industry should have at least 3 firms to make this adjustment. After the industry adjustment, we exclude stocks with a share price less than $5 or with a market
capitalization below the bottom NYSE size decile. The rest of the stocks are sorted into decile portfolios based on the goodwill proxies. Portfolios are rebalanced at the
end of each June. We report excess returns, three-factor alphas, and four-factor alphas. Newey-West adjusted t-statistics are reported in parentheses. The sample covers
1989 to 2016 because Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%. All returns and alphas are in percentage.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Low 1.29*** 1.25*** 1.17*** 1.10*** 1.13*** 1.07*** 1.05*** 1.03*** 1.12*** 1.05*** (4.37) (4.15) (3.63) (3.69) (3.63) (3.77) (3.58) (3.28) (3.96) (3.78)
High 0.48 0.57* 0.66** 0.79*** 0.69** 0.72** 0.66** 0.69* 0.80** 0.75** (1.52) (1.86) (2.17) (2.68) (2.25) (2.44) (2.02) (1.95) (2.56) (2.47)
Low − High 0.81*** 0.68*** 0.51*** 0.31** 0.44*** 0.35** 0.39** 0.34** 0.32*** 0.30*** (4.43) (4.66) (3.20) (2.30) (3.07) (2.56) (2.56) (2.21) (3.23) (2.98)
Three-factor alpha 0.86*** 0.71*** 0.48*** 0.26* 0.42*** 0.33** 0.39** 0.30** 0.35*** 0.32*** (4.33) (4.75) (2.79) (1.90) (2.78) (2.36) (2.44) (2.17) (3.70) (3.46)
Four-factor alpha 0.75*** 0.67*** 0.43*** 0.27** 0.38*** 0.33** 0.35** 0.32** 0.36*** 0.31*** (4.26) (4.48) (2.74) (2.03) (2.72) (2.45) (2.35) (2.09) (3.40) (3.28)
48
Table A3: Alternative Industry Adjustments to Goodwill-to-Sales
This table reports the equal-weighted average monthly excess returns and alphas to portfolios sorted on goodwill-to-sales with different industry adjustments. In column
(1), we use Fama-French 38 industries. In column (2), we do not adjust goodwill-to-sales by industry. In column (3), we use Fama-French 48 industries. In column (4),
we use Fama-French 30 industries. In column (5), we use Fama-French 17 industries. In column (6), we use Fama-French 5 industries. In column (7), we use 1-digit
SIC codes. In column (8), we use 2-digit SIC codes. The sample contains all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at
fiscal year-end, a price-per-share larger than $5, and a market capitalization higher than the bottom NYSE size decile. Following Fama and French (1992), we match
accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June of year t+1 to ensure that the accounting variables are known
before the returns they are used to explain. At the end of each June, we first compute goodwill-to-sale (GTS) as goodwill divided by total sales from the beginning of
the fiscal-year for all common shares traded in NYSE, AMEX and NASDAQ that have a positive goodwill at fiscal year-end. Then, we compute industry-adjusted GTS
(GTS_adj) as the difference between GTS and the mean GTS from the industry. We require that each industry should have at least 3 firms to make this adjustment. After
the industry adjustment, we exclude stocks with a share price less than $5 or with a market capitalization below the bottom NYSE size decile. The rest of the stocks are
sorted into decile portfolios based on industry-adjusted goodwill-to-sales. Portfolios are rebalanced at the end of each June. We report excess returns, Fama-French
three-factor alphas, and Fama-French-Carhart four-factor alphas, respectively. Newey-West adjusted t-statistics are reported in parentheses. The sample covers 1989 to
2016 because Compustat starts reporting goodwill in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in percentage.
(1) (2) (3) (4) (5) (6) (7) (8)
Low 1.29*** 1.11*** 1.18*** 1.27*** 1.37*** 1.08*** 1.11*** 1.08*** (4.37) (3.98) (3.39) (4.04) (4.72) (3.38) (3.62) (3.54)
High 0.48 0.56* 0.72** 0.55* 0.48 0.45 0.48 0.53* (1.52) (1.71) (2.47) (1.72) (1.50) (1.41) (1.58) (1.69)
Low − High 0.81*** 0.55*** 0.46** 0.72*** 0.89*** 0.63*** 0.63*** 0.55*** (4.43) (3.74) (2.22) (3.75) (5.57) (3.39) (3.80) (3.36)
Three-factor alpha 0.86*** 0.51*** 0.46** 0.78*** 0.96*** 0.66*** 0.61*** 0.55*** (4.33) (3.59) (2.41) (4.14) (5.43) (3.28) (3.32) (3.29)
Four-factor alpha 0.75*** 0.45*** 0.45** 0.71*** 0.84*** 0.60*** 0.55*** 0.55*** (4.26) (3.16) (2.30) (3.83) (5.22) (3.31) (3.20) (3.19)
49
Table A4: Alternative Factor Models and DGTW Adjusted Returns
This table presents equal-weighted portfolio alphas on Fama-French (2015) five factors, Fama-French-
Carhart six factors, Hou-Xue-Zhang (2015) q-factors, Stambaugh-Yuan (2016)’s mispricing factors,
and DGTW adjusted portfolio returns. We report alphas and adjusted returns for the bottom and the top
goodwill-to-sales decile, as well as a trading strategy that buys stocks from the bottom decile and shorts
stocks from the top decile. The sample contains all common shares traded in NYSE, AMEX and
NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5, and a market
capitalization higher than the bottom NYSE size decile. Following Fama and French (1992), we match
accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year t to June of
year t+1 to ensure that the accounting variables are known before the returns they are used to explain.
At the end of each June, we first compute goodwill-to-sales (GTS) as the ratio of goodwill to total sales
from the beginning of the fiscal year for all common shares traded in NYSE, AMEX and NASDAQ
that have a positive goodwill at fiscal year-end. Then, we compute industry-adjusted GTS (GTS_adj)
as the difference between GTS and the mean GTS from the industry. We use Fama-French 38 industry
classifications, and require that each industry should have at least 3 firms to make this adjustment. After
the industry adjustment, we exclude stocks with a share price less than $5 or with a market capitalization
below the bottom NYSE size decile. The rest of the stocks are sorted into decile portfolios based on
GTS_adj. Portfolios are rebalanced at the end of each June. Newey-West adjusted t-statistics are
reported in parentheses. The sample covers 1989 to 2016 because Compustat starts reporting goodwill
in 1988. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas
are in percentage.
Low High Low − High
Five-factor alpha 0.20 -0.51*** 0.71***
(1.31) (-2.72) (3.72)
Six-factor alpha 0.30** -0.36** 0.66***
(1.98) (-2.57) (3.58)
q-factor alpha 0.36** -0.38 0.74***
(2.04) (-1.61) (3.79)
m-factor alpha 0.40*** -0.28*** 0.68***
(2.70) (-2.84) (3.63)
DGTW Adjusted Returns 0.26** -0.51*** 0.77***
(2.40) (-4.29) (4.97)
50
Table A5: Independent Double Sorting Based on Institutional Ownership, Idiosyncratic
Volatility, and Market Capitalization, 1989-2016
This table reports the equal-weighted average monthly excess returns and alphas to three sets of
independently double sorted portfolios. In Panel A, we double sort our sample on industry-adjusted
goodwill-to-sales and institutional ownership. Institutional ownership is defined as the sum of
percentage holdings by institutional investors with 13F fillings. In Panel B, we double sort our sample
on industry-adjusted goodwill-to-sales and idiosyncratic volatility. We compute idiosyncratic volatility
as the standard deviation of the residuals from regressing daily returns on daily Fama-French three
factors for the previous year. In Panel C, we double sort our sample on industry-adjusted goodwill-to-
sales and market capitalization. For simplicity, we only report equal-weighted average monthly excess
returns and alphas for corner portfolios. The sample contains all common shares traded in NYSE,
AMEX and NASDAQ that have a positive goodwill at fiscal year-end, a price-per-share larger than $5,
a market capitalization higher than the bottom NYSE size decile. Following Fama and French (1992),
we match accounting data for all fiscal year-ends in calendar year t−1 with the returns for July of year
t to June of year t+1 to ensure that the accounting variables are known before the returns they are used
to explain. At the end of each June, we first compute goodwill-to-sales (GTS) as the ratio of goodwill
to total sales from the beginning of the fiscal year for all common shares traded in NYSE, AMEX and
NASDAQ that have a positive goodwill at fiscal year-end. Then, we compute industry-adjusted GTS
(GTS_adj) as the difference between GTS and the mean GTS from the industry. We use Fama-French
38 industry classifications, and require that each industry should have at least 3 firms to make this
adjustment. After the industry adjustment, we exclude stocks with a share price less than $5 or with a
market capitalization below the bottom NYSE size decile. The rest of the stocks are sorted into quintile
portfolios based on GTS_adj. We further sort these stocks by terciles independently based on
institutional ownership (Panel A), idiosyncratic volatility (Panel B), market capitalization (Panel C),
and compute the equal-weighted returns for the 15 (3×5) double sorted portfolios. Portfolios are
rebalanced at the end of each June. We report excess returns, Fama-French three-factor alphas, and
Fama-French-Carhart four-factor alphas, respectively. Newey-West adjusted t-statistics are reported in
parentheses. The sample covers 1989 to 2016 because Compustat starts reporting goodwill in 1988. *,
**, and *** indicate significance at 10%, 5%, and 1%, respectively. All returns and alphas are in
percentage.
Panel A: Institutional Ownership High IO Low IO
Low
GTS_adj
High
GTS_adjLow − High
Low
GTS_adj
High
GTS_adj Low − High
Excess return 1.15*** 0.88*** 0.27** 1.25*** 0.16 1.09*** (3.99) (3.04) (2.26) (3.56) (0.46) (5.16)
Three-factor alpha 0.14 -0.21 0.35** 0.30 -0.85*** 1.15*** (0.98) (-1.32) (2.50) (1.45) (-4.74) (5.40)
Four-factor alpha 0.14 -0.07 0.21 0.49*** -0.58*** 1.07*** (1.07) (-0.47) (1.60) (2.72) (-2.65) (4.73)
(Continued)
51
(Continued)
Panel B: Idiosyncratic Volatility High IVOL Low IVOL
Low
GTS_adj
High
GTS_adjLow − High
Low
GTS_adj
High
GTS_adj Low − High
Excess return 1.18*** 0.11 1.07*** 1.11*** 0.92*** 0.19* (3.03) (0.24) (5.04) (5.33) (3.50) (1.66)
Three-factor alpha 0.05 -1.06*** 1.11*** 0.29** 0.01 0.28*** (0.33) (-5.41) (5.17) (2.29) (0.07) (2.65)
Four-factor alpha 0.21 -0.71*** 0.92*** 0.32*** 0.09 0.23** (1.23) (-3.90) (4.45) (3.16) (0.86) (2.37)
Panel C: Market Capitalization Big Firm Small Firm
Low
GTS_adj
High
GTS_adjLow − High
Low
GTS_adj
High
GTS_adj Low − High
Excess return 1.09*** 0.84*** 0.25** 1.24*** 0.53 0.71*** (4.66) (3.15) (2.41) (3.86) (1.49) (4.12)
Three-factor alpha 0.17 -0.15 0.32*** 0.22 -0.55*** 0.77*** (1.33) (-1.18) (3.12) (1.51) (-4.29) (4.93)
Four-factor alpha 0.28** 0.02 0.26** 0.32** -0.42*** 0.74*** (2.27) (0.17) (2.44) (2.41) (-3.14) (4.79)
top related